Statistical Data Analysis based on Linked Open Data

Abstract

Research in the social sciences is based on the analysis of social phenomena via quantitative evidence. Scientists typically need to perform major and complex analyses on statistical data, but as part of their main tasks, they also require tedious secondary examinations on heterogeneous and distributed datasets, for example to verify prior or referenced assumptions or detecting correlations between two or more datasets. A lot of tools already exist which support researchers in processing and analysing their data effectively, but raw data often has to be converted to particular formats in order to be processed and analysed. In this paper, we propose a method to perform such statistical analyses on Linked Open Data resources in order to support common tasks that researchers encounter when working with heterogeneous and distributed datasets. The idea of Linked Open Data provides a technical basis for exposing, sharing and linking data on the web, based on an established web architecture, comprising standardised formats and interfaces. However, statistical calculations cannot be performed directly on these data sources yet. Our prototype covers some common exemplary tasks for data analysis in the social sciences.